Measurement Needs for Understanding the Changing Nature of Work
Driven both by business needs and by the desire of workers for a broader set of work options, work arrangements have changed dramatically in recent decades. External factors having to do with the macroeconomy, such as fallout from the Great Recession, and changing technology have also played roles in shaping the emerging models of employment. Internet platform-mediated work, at this point only a small component of the overall labor market, is but one example of the flux now being experienced. Better data are needed to fully understand these ongoing labor market transitions and to inform new policies being called for to enhance the positive aspects of these changes and mitigate the negative impacts. Specifically, data are needed to support analyses connecting changes in business structure and employment relationships with job and worker outcomes.
In this chapter, we discuss the changing nature of jobs—whether it is the contingency emphasized in the early Contingent Worker Supplement (CWS) to the Current Population Survey (CPS) or the broader job security and employee protection issues that have emerged since—along with the measurement implications. We identify the kinds of information about modern work arrangements needed for policy and research purposes and consider the priorities for collecting this information. We discuss the scope of work and income sources that are important to measure (section 2.1); how different categories of work factor into policy strategies (section 2.2); and, most importantly, how job characteristics affect worker outcomes and well-being, as well as employer hiring practices (sections 2.3 and 2.4). The strengths and shortcomings of existing data sources in meeting these information needs are also noted. Those sources include but are certainly
not limited to the CWS. We leave it to subsequent chapters, however, to build on the conceptual discussion here and advance recommendations for improving measurement capacity—both through a revised CWS (Chapter 3), and by building on other, potentially complementary data sources (Chapter 4).
Periodic, cross-sectional household datasets such as the CWS can shed light on some important policy questions related to alternative work arrangements (AWAs):
- Among people engaged in AWAs, what is the relationship between worker age, education, and other characteristics (gender, race/ethnicity, immigration status) and job opportunities?
- Why do workers participate in AWAs? To what extent does participation reflect worker preferences for flexibility or other AWA attributes, and to what extent would those in AWAs prefer a standard employment arrangement?
- Is engagement in AWAs more likely among multiple job holders or as a secondary work activity? And what is the impact of multiple job holding on worker well-being and on work/life balance?
For other research and policy questions, panel datasets or (as discussed in Chapter 4) high-frequency nonsurvey data, such as commercial and administrative data, can be instrumental in helping to understand emerging labor force dynamics. Examples of such questions include these:1
- Does participation in AWAs increase (due to personal financial needs) or decrease (due to reduced demand for services) during economic downturns?
- How do AWAs help workers smooth their income and consumption over time?
- What are the career paths that lead into and out of AWAs?
Recognizing the value of longitudinal data collection, Chapter 4 discusses the value of adding job questions to surveys such as the Census Bureau’s Survey of Income and Program Participation (SIPP) and the National Longitudinal Surveys.
1 Although the CWS is a supplement to the CPS, which has a limited longitudinal structure, researchers have used the CWS to examine some of the employment dynamics associated with AWAs. For discussion on using the limited longitudinal structure of the CPS together with the CWS to study the stability of AWAs, see Addison and Surfield (2009) and Houseman and Polivka (2000). For discussion of worker paths into AWAs, see Addison and Surfield (2006) and Farber (2017).
Other kinds of questions may be best addressed by analyses using business-sourced data:
- What kind of firms use AWAs and why?
- How have the activities and occupations where AWAs are used changed over time?
- What determines the diffusion of AWA usage in an industry and occupation?
- How does growth in the use of AWAs relate to the increasing separation of firms between high-productivity / high-wage and low productivity / low-wage firms?
- Are AWAs a first-order result of employment restructuring or a second-order effect of business restructuring?
- How do businesses view their relationship with AWA workers? Do they think of themselves as traditional employers? Indirect employers? Supply contractors?
- To what extent is the use of AWAs a reflection of more efficient ways to organize production, given new technology, versus a method to thwart workplace protections and social insurance protections?
Finally, the changing business structures affecting employment arrangements come into play, at least indirectly, for an even broader set of social and economic questions affecting the well-being of individuals and households:
- What role, if any, is the trend toward AWAs playing in polarizing high-skill and low-skill jobs (and the disappearance of “solid” middle class jobs) and in the growth of earnings inequality?
- What is the relationship between AWAs and the trend toward declining labor mobility and opportunities for career progression?2
- Is the changing structure of work related to declining labor force participation among certain demographic groups (e.g., white males)?3
- Do AWAs draw people into the workforce who would otherwise be nonparticipants?
- Are certain types of AWAs replacing other types?
2Davis and Haltiwanger (2014) provide empirical evidence for a decline in labor market mobility.
3 If the earnings, benefits, and conditions of work associated with an occupation or industry decline as a result of AWA utilization, the pool of workers in those jobs could decrease, affecting labor force participation for particular segments that were once attracted into the workforce by those types of jobs.
Addressing these questions will require many kinds of data and involve many research threads. Nevertheless, if we are to gain a better understanding of the future of work, development of the right kinds of data instruments is crucial.
2.1. EMPLOYMENT: MEASURING ALL “SIGNIFICANT” SOURCES OF WORK INCOME
Which Work Activities Are in Scope?
In its role measuring the nation’s labor market activities most pertinent to research and policy, BLS cannot attempt to capture all sources of income. The agency’s focus rightly is on income generated from market work. Work activity associated with individuals’ primary jobs therefore is clearly within scope, while work conducted for family or friends (e.g., household production) or for a neighbor in a barter-type agreement is just as clearly out of scope.4 But much work that routinely takes place—ranging from selling used items on eBay or at a flea market to earning income from working a second or third job, perhaps sporadically—falls between these extremes. Currently in the CWS, deciding what to include as “work for pay” is largely up to respondents’ interpretations.
AWAs are of course not new. Examples of inside contracting systems as well as networks of outside contractors could be found in 19th century U.S. factories. The growth of standard employment relations in the post–World War II period reflected efforts by employers to create strong internal labor markets wherein employers could invest in workforce training and expect to retain those human capital investments. This made sense for employers during the three decades after World War II, as the economy grew and the strong position of companies in domestic markets allowed them to negotiate with unions in a way that increased wages and decreased wage inequality. Subsequently, the relationship between workers and firms weakened. Beginning in the mid-1970s, macro trends such as globalization and international competition, technological change, and the decline of unions led employers in the United States to seek the flexibility to more readily adjust the size of their workforces (National Academies of Sciences, Engineering, and Medicine, 2017).
4 Several surveys—such as the Federal Reserve Board’s Survey of Household Economics and Decisionmaking, described in Chapter 4—are oriented to pick up informal work. Specialized surveys—for example, a national-level household survey of informal work conducted by Jensen, Tickamyer, and Slack (2019)—have even included questions about informal work that is done on a barter or other nonmarket/nonmonetary basis. Some of these surveys indicate relatively small amounts of income being generated, but very high participation rates, in informal work.
Temporary agency employment grew rapidly during the 1970s and continued growing all the way through the 1990s, although it still represented a relatively small portion of the overall labor force. As described in Chapter 1, by the time the BLS sponsored the first Contingent Work Survey in 1995, trends in the temporary help services industry, and in outsourcing generally, were readily apparent. However, the share of the workforce engaged in temporary help agency work, as measured in the CWS, has been fairly stable over the survey’s history, including the most recent 2017 wave.
Since 1995, other forms of work have evolved and grown in relative importance, garnering the attention of policy makers and researchers. More prominent in today’s economy than temporary help agency work—and more salient from the point of view of policy—is independent contracting, which includes platform work. Under this arrangement, individuals are self-employed rather than being employees of the organizations or individuals for which they perform work. Of the AWAs measured in the CWS, independent contracting is the largest group, although the percentage of all employment accounted for by this category as measured in the CWS changed little between 1995 and 2017.
As described in detail in Chapter 3, however, the CWS estimates likely underestimate the proportion of persons who work as independent contractors. In large part this is because the survey asks only about a person’s main job, and many people work as independent contractors on a second or third job. Research evidence also suggests that many people working solely in these nonemployee arrangements are not captured in household surveys like the CWS. The independent contractor group is critical for policy, as these individuals typically cannot rely on employing organizations for benefits and security, and they are not covered by employment and labor laws that provide basic protections to workers (such as minimum wages) and access to social insurance programs (such as unemployment insurance and workers’ compensation).
One question of appropriate measurement scope that comes into play conspicuously in observing independent contracting concerns the distinction between “work-based” income and income derived from a combination of work and capital-generated activity. As quickly becomes clear in thinking about some of the newer platform companies, the line between these two activities can sometimes be blurry. For example, renting out rooms through AirBnB often is thought of as deriving income from a capital asset, but it also has a work- (time-) based component; in fact, a person operating several AirBnB properties may be occupied full-time in the enterprise. On the other hand, even ride-share workers, ordinarily considered to be engaged in labor-platform work, must have capital in the form of a car that passes
the company’s equipment standards.5 It is an open question how far apart on the labor/capital spectrum the work is through platforms such as, say, AirBnB and Uber.
There is a conceptual arbitrariness about drawing the line of inclusion for measurement in labor statistics somewhere between the two kinds of jobs, labor- versus capital-based. Furthermore, this distinction is a problem that is pervasive for all types of self-employment, not just AWAs. Income from traditional self-employment also may combine returns to labor and capital. Further complicating measurement is the fact that, for most purposes, the relevant concept is net income (essentially what a person would report on a Schedule SE), but it may be difficult for households to separate out the costs associated with the work they did “last week” (a point reiterated later in this chapter). As discussed below, in the CWS, hours may be the best measure of the intensity of the activity.6 We address this issue of demarcating returns to capital and labor in later chapters, where we also discuss nonsurvey data approaches to measuring the extent of independent contracting and other self-employment work and its contribution to the economy.
In recent decades, business restructuring has affected the prevalence of AWAs across the economy, and in a way that extends far beyond the comparatively recent emergence of the internet platform-based work alluded to above. Profound organizational changes have found major businesses focusing on integrating available resources into their value creation process to best provide value to customers and investors; this often involves concomitantly contracting with other entities to carry out those efforts (Appelbaum and Batt, 2014). The organizations undertaking these activities for lead businesses are guided by exacting standards and high-powered incentives to ensure that core competencies are met. These take the form of detailed subcontracting and supply-chain requirements; franchise agreements; and, most recently, the highly calibrated incentive systems created by platform algorithms. Such restructuring includes but is not limited to the AWAs captured in the CWS.7
5 There is a growing list of jobs requiring a car that are clearly income-generating—Amazon Flex, Uber Works, Doordash, etc.—that are relevant to many of the preamble questions to this chapter. These platforms reflect activities previously done as work by workers (as opposed to the AirBnB example, which is less clear).
6 An option used in some data collections, such as the Survey of Informal Work Participation sponsored by the Federal Reserve Bank of Boston, is to take a “time for income” approach as opposed to the current work for income approach. Regarding platform work, an approach whereby respondents are asked whether they spent time providing services or selling goods for pay could be considered.
7Weil (2019) argues that the wider definition of “fissuring” implies that the part of the workforce affected by these changes is much larger than is implied by household measures of AWA.
Changes in the structure of employer-worker relationships in recent decades have given rise to a different set of individual and household choices. These changes require rethinking the meaningfulness of terms like “primary job” and “supplemental work” and of the factors contributing to decisions by people to take different kinds of jobs. Data sources such as the CWS are needed to better inform questions about how changes in primary job characteristics are affecting worker outcomes. Similarly, information is needed about the prevalence of additional work beyond the primary job—much of which is likely to be in AWAs—and the reasons, including financial reasons, for taking on additional work. In many cases, secondary jobs are critical sources of income for households, and thus information on all jobs is necessary to understand how people combine different work activities to earn a living. For these reasons, it is necessary for policy purposes to collect information about additional work beyond the respondents’ primary job. Because secondary work activities are often engaged in sporadically, the CPS reference period “in the past week” would not capture them comprehensively. This issue of the length of the reference period used in the survey and its potential impact on various estimates of AWA activity is discussed in Chapter 3.
It is not practical for a single survey (or nonsurvey data source for that matter) to generate a profile of the full characteristics of all work activities in which people engage. Information about hours and earnings on all jobs is essential. Additionally, it would be valuable and within scope for the BLS to include questions on why respondents have a second job. Policy discussions regarding implications for the well-being of people who engage in AWAs often make assumptions about workers’ underlying motivations and whether the AWA is their primary or a secondary job. For example, some independent contractors may prefer greater job flexibility and a desire to pursue entrepreneurial opportunities. This characterization may be most accurate when describing workers who seek out AWAs as their main source of income. In other cases, AWAs may reflect hard choices made by individuals and households to make ends meet, particularly when they are engaged in the work to supplement income earned in a traditional job.
Nevertheless, even when they are second jobs, AWAs may be an attractive means of fulfilling a desire for supplemental income or satisfying targeted savings goals. Probing the assumptions about preferences should be combined with mapping preferences against other worker characteristics, such as educational background, age, and experience, as well as job characteristics, such as occupation, skill requirements, and industry. A deeper understanding of the patterns of preference among workers taking on multiple jobs is critical to framing future policy choices. Achieving such
an understanding will require both accurate data and sound approaches to measuring the full range of market work in which people are engaged.
2.2. JOB TYPES: CATEGORIES OF ALTERNATIVE WORK ARRANGEMENTS
The CWS provides key measures of temporary (contingent) work and of the AWAs often associated with temporary work, with outsourcing in its different forms, and with unpredictable schedules. Disagreements exist, however, among researchers, policy makers, and other stakeholders about the definitions and measures of these concepts and priorities for future data collection.8 As previously discussed, the CWS collects information on the following five major types of AWAs:9
- Temporary agency work. In this case, the worker is an employee of the temporary help agency and is assigned to work for clients, typically at the client’s worksite. As discussed in Chapter 3, employer survey data indicate that the share of the workforce in temporary agency employment is understated in the CWS. In May 2017, according to the CWS, about 1.4 million workers (0.9% of total employment) were paid by a temporary help agency.10
- On-call work. On-call work is a type of on-demand work in which the worker is called to work at the job only when needed. In on-call work, like other types of on-demand work, the number or the timing of hours worked varies and, importantly, the hours are controlled by the employer, not the employee. These characteristics are not necessarily unique to on-call work. In May 2017, according to the CWS, there were 2.6 million on-call workers (accounting for 1.7% of total employment).
- Independent contracting. Independent contracting is a type of self-employment. The term typically refers to the work of those who are self-employed and do not own or run a business or do not have a sizable capital investment in a business. Independent contractor arrangements are often complex, involving tiers of subcontracting that may lead to misreporting and undercounting on tax forms. Independent contractors also include most types of work on platforms and other informal, nonemployee work arrangements. As
8 For a detailed discussion of these definitional issues, see Allard and Polivka (2018).
9 The CWS also collects information on day laborers but, given their small number as captured in the survey, the BLS does not report this category in its data summaries.
10 This figure, and the parallel estimates for other AWA categories, are available: https://www.bls.gov/news.release/pdf/conemp.pdf.
- is discussed in Chapter 3, research evidence also suggests that independent contractors are underreported in household surveys. In May 2017, according to the CWS, there were 10.6 million independent contractors (6.9% of total employment).
- On-demand, platform-intermediated work. Platform-intermediated work, sometimes referred to as “gig” work, may cover personal service activities, such as child care, house cleaning, or ride sharing, as well as goods-related activities, such as selling goods online or renting out property—though each of these activities is often not mediated by a platform. As noted above, many people who engage in platform work use it to supplement their income, but some rely on it as their main source of income. Some platform-intermediated activities are done occasionally and do not take much time, and thus may not fit neatly into a standard concept of what is considered to be “work” (see Abraham and Amaya, 2019; Farrell, Greig, and Hamoudi, 2018). In some cases, internet intermediary companies have formalized the arrangements for certain types of work previously done on an informal basis. For example, a neighborhood dog walker who previously had operated by word of mouth might now use Rover.com, or a person who had previously rented out a spare room through classified advertising might now use AirBnB. For May 2017, BLS estimated that about 1.0 percent of total employment was accounted for by electronically mediated workers.
- Contract company work. This category refers to workers, besides temporary agency workers, who are employees of a company but whose services are contracted out to clients. As discussed in Chapter 3, it can be difficult to measure this category of work because it is unclear how accurately contract company workers are able to report their status as such. The narrow BLS definition may capture that particular part of contract company work with reasonable accuracy—those who work for a company that contracts out their services and who primarily work for one client at the client’s worksite—but it misses much of the network of contract activity that is also of interest. In May 2019, according to the CWS, 933,000 workers (or 0.6% of total employment) were engaged in contract company work.
Even with a blurring of work categories, a key measurement objective remains to track whether workers are employees or independent contractors. The policy relevance here is obvious. By virtue of the fact that they are self-employed, independent contractors are not covered by a host of worker protections, including minimum wage, overtime pay, health and
safety rules, sexual harassment laws, and rights to organize and collectively bargain, all of which apply only to employees. Nor are they covered by social insurance benefits, such as workers compensation and unemployment insurance, or by employer-provided benefits, such as health insurance and retirement plans.
In addition, the earnings of an independent contractor, net of expenses, may in some cases be substantially below those of workers undertaking similar work as employees. For example, it has been estimated that when fully accounting for vehicle fuel, amortization, insurance, maintenance, tolls, and other costs, drivers working as independent contractors for Amazon Flex (the next-day delivery arm of Amazon) received estimated net earnings of $5.30 per hour—significantly below the federal minimum wage. This compared to average earnings of $23.10 for UPS drivers and $14.40 for FedEx drivers (Vernon, 2018; also see Zaleski, 2018). Similar outcomes may arise in certain franchise relationships prevalent in industries like janitorial services and home care, where the franchise agreement may put the franchisee in a situation akin to employment rather than running a business (Weil, 2014, Chapter 6). Higher rates of violations in health and safety, labor standards, and other workplace requirements have also been documented for those in subcontracted or franchised relationships.11
The implication of these differences between employees and independent contractors is that, if work continues to migrate toward a nonemployer structure, there will be a need to adjust social safety net and employment laws in order to mitigate the potential negative side effect of this trend while realizing its potential benefits. From a policy perspective, it is critical to capture the distinction between employees and nonemployees. The current CWS attempts to do this by asking whether respondents are independent contractors, independent consultants, or freelance workers, but the wording of the question may not pick up everyone who is working but not as an employee. We return to this issue in Chapter 3 to recommend a modification to the wording of relevant questions.
Another data collection implication of these complexities in organizational and legal relationships, and of possible misunderstandings by workers concerning their status, is that administrative data can provide an important complement to survey data for measuring self-employment, including independent contractor work. As discussed in Chapter 4, it also may be possible to advance the measurement of independent contract work using combinations of survey and nonsurvey (administrative) data.
A final point about categorizing types of work is that, along with the evolution of AWAs, standard work arrangements also have experienced
11 See, for example, Grabell, Larson, and Pierce (2013); Jamieson (2014); and Ji and Weil (2015).
change. Relative to the more manufacturing-based economy of the past, many employer-based jobs now offer fewer benefits (e.g., pensions, health care) and may offer less stability. Studies have also documented reductions in earnings and benefits for workers whose jobs were subcontracted to third parties. These developments serve to lessen the distinction between, say, W-2 jobs and some of the AWA categories listed above.
Reiterating the discussion above, a critical distinction for policy discussions about work arrangements is whether a person is considered an employee or not. Among the self-employed, it also is useful to know if the individual owns or operates a business, or if the individual is an independent contractor or other worker in a nonemployee arrangement. Among employees, it would be valuable to know whether the worker is in a bilateral or trilateral employment relation. Bilateral employment relations encompass standard employment relations as well as direct-hire temporaries. Trilateral employment relations involve those who work in intermediated arrangements, such as for temporary help agencies, leasing agencies, or other contract companies. Including descriptions of different types of workplace intermediaries would also be a useful path forward and allow some continuity with earlier CWS categories, while acknowledging the problems detailed in this section. The important types of distinctions that would be missed under such a realignment include franchised relationships and subcontracted (and possibly third-party managed) arrangements, but those can be more accurately measured through business-focused surveys.12
2.3. KEY JOB CHARACTERISTICS AFFECTING WORKER OUTCOMES AND WELL-BEING, EMPLOYER STRATEGIES
The primary research and policy goal motivating the collection of data about AWAs in the CWS has been to understand the nature of the work, how it is changing, and the implications for worker (and employer) outcomes. Most AWA measurements in the CWS were intended to capture situations in which workers are not employees of the organization using their labor and where there is a weakening of the attachment between workers and firms (as in firms’ use of temporary help workers, independent contractors, contract company workers, or day laborers). The CWS also measures other work arrangements that are characterized by a high level of
12 The specific type of subcontracting referenced here is similar to outsourcing: It reflects a decision by a business to take some activity and purchase it as an outside service rather than doing it internally. The work itself could still be done by W-2 workers in the business the subcontracted work is going to (or it might be to a business that is hiring its workers on a 1099 basis). The point is that this is work that has been affected by fissuring but would not be classified as an AWA even though the impacts of such arrangements still need to be measured for the reasons discussed in the report.
precarity, such as work that is temporary or on-call. Yet work arrangements are not always easily labeled. Terminology may not be well understood by many respondents, and it may change over time.
The nature of “employment” itself has become complicated because of the impact that this business restructuring has had on the employment relationship. Even “employee” is a contested concept. Workers may not always know who their employer(s) is (are) because of the multitiered, fissured nature of business relationships and legal complexities. For similar reasons, they may not know if they are actually employees or independent contractors (self-employed).13 Independent contractors may accurately report working for an employer, particularly if the worker is reliant on one business for work (independent contractors are not necessarily hustling for clients).
Given the goals of the CWS and other BLS data programs, a reasonable strategy would be to place less emphasis on questions asking respondents to classify their job into one of several categories and, instead, ask questions that elicit information on the organizational arrangement(s) under which people work, specifically job characteristics and work outcomes. This section identifies the key job characteristics and work outcomes on which measurement should focus and clarifies why that focus matters. The key job characteristics are hours, scheduling variability, and contingency. The key work outcomes are earnings, benefits, and workplace safety.
Hours and Scheduling
To support the nation’s basic employment statistics, there is clear value in collecting information on total hours from a main job and, in cases where a person has more than one job, total hours from all employment. As pointed out above, however, variability of schedules and reliability of hours are also of considerable interest, given their relationship to job and economic insecurity (Henley and Lambert, 2014).14 Collecting this kind of information requires going beyond asking respondents only about a “typi-
13 As discussed in Chapter 3, there is also an important distinction between whether people know if they are independent contractors or employees and whether the questions on the CPS/CWS elicit accurate responses. One conclusion of Abraham, Hershbein, and Houseman (2019) is that the latter is a problem, too. Household surveys, including the CWS/CPS and American Community Survey, do not ask whether an individual is an employee—only whether they work for an employer/organization.
14 Indicative of the current policy interest in unpredictable work schedules is U.S. Senator Elizabeth Warren’s inclusion of the issue (her “Fair Workweek” plan to give workers more advanced notice of their schedules) as part of her presidential platform. Available: https://elizabethwarren.com/plans/part-time-workers.
cal week.” Chapter 3 proposes some alternative approaches to generating data that capture the extent of scheduling and income variability.
When work hours are unpredictable and variable, it is important to know who sets the schedules—Is it the worker or the employer?—and with how much advance notice. A self-employed person who chooses to work 20 hours a week represents a very different scenario from a person who is limited to the same number of hours but would prefer to work more to meet household economic needs. The policy implications are clearly different if the trend toward irregular or reduced hours is being driven by workers’ desire for flexible scheduling as opposed to being driven by lack of opportunities to acquire “steady work.” In any case, for workers who seek flexible work, flexible scheduling may not be an inferior arrangement.
To take an example, it might turn out that many jobs are characterized by week-to-week variability in hours. The growth of scheduling algorithms in retail and other service sectors, where workers may not receive much notice about work schedules, suggests that the “predictability” distinction/advantage of W-2 work may be eroding (Henley and Lambert, 2014). It is therefore important to collect data on and measure variability in hours for all but self-employed workers. As discussed in the next chapter, because of the growing importance of these issues, yielding room in the survey from the old contingent-work questions and providing additional questions on worker preferences regarding hours may be a worthwhile tradeoff.
An essential aspect of work arrangements measured using data from the CWS is the “contingency” or, perhaps more accurately, the temporariness and insecurity of certain jobs. As noted at the outset of this chapter, the need to measure the perceived expansion of temporary work drove the specifications of the CWS initially. The survey is well constructed to pick this up, with the caveat that it can be difficult for individuals to accurately report how temporary their job is. Compared to many other countries, however, the percentage of workers identified as “temporary” is relatively low in the United States. This renders the category less important than some others, such as independent contracting.15
The characteristic of temporariness is only one factor contributing to the sense of job security a worker feels. Job security is more complex than simply having a job or not; moreover, rather than being defined as either secure or insecure, jobs have aspects that make them more or less secure in certain respects. For example, many jobs—including many standard
15 In other countries, temporary workers often have explicitly temporary contracts, so that they are a clearly identifiable category; in the absence of these types of contracts in the United States, the distinction is less clear-cut.
employer-employee jobs, especially in retail and other service sectors—have an on-call dimension that leads to high variability in the timing and even quantity of hours of work. Work variability of this variety leads to a different kind of job insecurity than does temporary work, but nonetheless it is important to measure and understand.
People do not think uniformly about temporariness, and it is an open question how relevant this job dimension still is to policy. Viewed from a positive perspective, temporariness means greater employment fluidity, and for some people new temporary work arrangements have added a welcome dimension of flexibility to their earning activities. Technologies have emerged in the past decade that enable workers to earn extra money on the side simply by turning on an app, such as Uber (or any of at least 128 others16), when they have a free hour or two. This means that some workers, such as those whose preferred approach is to pursue a variety of short-term jobs, are predisposed to be less concerned about the longevity, or “security,” of a job. By their nature, these jobs are not permanent in the conventional sense; they are flexible. If an Uber driver works for a month, then does something else, and then returns to Uber driving later in the year, the concept of temporariness may not be terribly relevant—or at least, in that worker’s thinking it may not be an automatic negative.
On the other hand, for workers who would prefer a more stable commitment by a job provider, temporariness may well be considered a negative to their sense of security and well-being. How, then, should job insecurity or reliability be measured to take into account these divergent impacts on outcomes? Because of these nuances in the ways people view job flexibility and job security tradeoffs, it is important to gain a better understanding of worker preferences regarding the kinds of work in which they engage or that they are pursuing.
An additional concern arises if, as some evidence suggests, all work has increased in precarity over time and workers’ ability to assess insecurity is limited and affected by changing expectations (Howell and Kalleberg, 2019; Kalleberg, 2011). The approach of identifying preferences in work type is one useful way to get at this, and it could certainly be pursued in household surveys like the CWS. Since many jobs (both standard and AWA) are likely to be insecure and hence “temporary” to some extent, a potentially useful approach to measuring the insecurity aspect of work arrangements is to directly ask respondents about the likelihood that they will lose their job and, if they did lose it, how difficult it would be to find a new one. These approaches, used in several European surveys, are discussed in Chapter 3.
16Farrell, Greig, and Hamoudi (2018) identify 38 million payments directed through 128 different online platforms to 2.3 million distinct Chase checking accounts.
As explicitly recommended in Chapter 3, the BLS should continue to collect data that helps researchers understand why people engage in multiple jobs—including why they pursue platform work on the side. Questions on motivation for working a second job (work activity) are crucial for understanding people’s work patterns. Relatedly, questions about job satisfaction for all workers may also provide insights into how respondents view the relative desirability of different work arrangements.
The positive relationship between people’s earnings and their well-being is obvious. Much attention has been given to the lack of growth in earnings experienced by workers over recent decades and the negative economic, health, and social impacts this has had on families. But how does this relate to changing work arrangements, an expansion of AWAs, and the potential ripple effects back on the conditions in standard employment settings? Whether changes in workers’ employment arrangements account for the growth in earnings inequality, for example, is an open question.
Recent research (e.g., Howell and Kalleberg, 2019) argues that the growth of AWAs does not account for the decline in job quality, which is due more to the decline in the quality of standard jobs. This distinction, however, is difficult to make, especially when there is interdependence between AWAs and employer jobs. An important issue for further study is whether AWAs also create spillover effects that influence earnings and conditions in traditional work relationships. Additionally, some AWAs are not well identified in the existing data and may appear as traditional employment arrangements. Outside of certain industries, like temporary help, it is notable that contract work, outsourced work, and other fissured work arrangements are simply not measured, so workers in those arrangements may appear to be in regular wage and salary employment.
It is difficult to generalize about differences in earnings between workers in different arrangements. For example, some highly skilled independent contractors in high tech and other professional occupations earn higher wages than their counterparts in standard full-time jobs, although they are still less likely to have fringe benefits (Kalleberg, 2011; Kalleberg, Reskin, and Hudson, 2000). On-call workers, day laborers, and part-time workers, however, tend to earn consistently lower pay and have fewer benefits; except for part-time workers, they also tend to express a preference for not working in these jobs (Barley and Kunda, 2006).
To accurately compare people working in independent contracting arrangements with those working in wage and salary jobs, due to the
expenses connected with self-employment it is important to make a distinction between gross earnings and net earnings.17 An example is fuel and auto depreciation for platform drivers and delivery workers. As discussed in Chapter 3, the CWS ideally would collect information on expenses so that both gross and net earnings could be reported. In practice, however, it is likely that many independent contractors would find such information difficult to report. While it would not be possible to ask occupation-specific questions in the CWS, specialized surveys have been used to probe into some jobs—such as those connected with driving or auto ownership, where the dominant platform models rely on cars for their production process—to follow up with questions about expenses. Such a question, for example, might ask, “On average, how many miles did you drive for this work?” or “How long have you owned the car for which you undertake this work and how much did you pay for it?”18
Jobs in various work arrangements may also systematically differ not just in terms of the level of earnings but also in the variability of earnings. In the CWS, earnings variability could be captured indirectly by questions about variability of hours. As with predictability and reliability of work hours, predictability of earnings has direct implications for people’s economic security.
Given the employer-delivered nature of many U.S. social welfare protections, workers in alternative arrangements are also likely to be disadvantaged with respect to retirement plans, medical care plans, paid time off, and other benefits. This is another case in which it is important to delineate information about primary jobs and secondary jobs and determine whether the worker has benefits from any job or through the job of someone else in the household.
Data are needed to inform policy initiatives, such as the recently passed California legislation known as AB 5,19 which seeks to help miti-
17 Asking tax-related questions carries with it the danger of potentially harming survey response rates and accuracy. One alternative source that researchers (e.g., Collins et al., 2019; Lim et al., 2019) have explored for such information on contract workers is tax returns. A perfectly completed tax return (Schedule C) subtracts costs from the gross earnings reported on, say, a 1099-K form. A survey could ask respondent if they had expenses, and then, if data linking were possible, these records could triangulate to tax data. This possibility is further developed in Chapter 4, which underscore how administrative tax data could complement data collected in household surveys.
18 See for example: https://www.ridester.com/2018-survey.
19 The AB 5 law tightens the definition of employment and is intended to reduce the misclassification of employees as independent contractors—common in such sectors as ride-hailing drivers, construction workers, food-delivery couriers, nail salon workers, and franchise owners—in an effort to reduce the insecurity associated with this kind of work.
gate the negative side effects being created by shifts in certain sectors toward increased use of independent contracting. Basic information on the incidence of benefits being offered, even if it is generated from data without much nuance on the level of benefits, would be useful for these discussions.
Other Job / Work Outcomes
Workplace health and safety can be affected by factors linked to AWAs. An analysis of the Census of Fatal Occupational Injuries found that, in 2017, about 12 percent of fatal workplace injuries were experienced by independent workers (defined as workers with short-term jobs that involve a discrete task and have no guarantee of future work). This represents a disproportionally higher propensity of death attributable to a workplace incident than that experienced by their employee counterparts (Pegula and Gunter, 2019). Health and safety risks arising from fissured employment relationships can also spill over to other parties; for example, Litwin, Avgar, and Becker (2017) find that outsourcing hospital cleaners increases the spread of health care-associated infections. As noted above, employment protections, rights, and social safety net protections may be negatively affected for workers who are in AWAs (e.g., have unstable hours or are on-call) and, if there are spillover effects, even for workers who are designated as employees.
The CWS cannot be the data source to provide statistics about all of the above as they relate to work arrangements. Data collection will need to be prioritized according to the value the various types of data have in addressing policy issues as well as what it is feasible to collect on a household survey. Among the most important kinds of data are those that illuminate the following:
- Distinctions of employee versus self-employed/independent contractor;
- Categories of AWAs, such as temporary agency work and platform-based work;
- Job characteristics, starting with earnings and variability in earnings (which the CWS is well suited to address), for primary and secondary jobs;
- Multiple job holding (primary and one secondary job);
- Control over work schedule;
- Health insurance support; and
- Preferences for AWA versus more standard work arrangements.
2.4. INFORMATION ABOUT ALTERNATIVE WORK ARRANGEMENTS THAT COULD BE PROVIDED BY BUSINESSES
Underlying structural relationships have emerged that affect work and worker outcomes, and the CWS was created to study the role of AWAs in these changes. Workers may not be in the best position, however, to report on the sometimes complex work arrangements they have.20 For example, workers often are employed by firms that are in contract or subcontract arrangements with other firms that, research suggests, may impact the worker outcomes. The CWS only tries to measure workers whose employers contract out their services if they primarily work for one client at the client’s worksite; otherwise, the contract relationship is deemed too complex to ask about in a household survey.
Information from business surveys, ideally linked with information from employee surveys or administrative data, is needed. In many of these situations, some work that was once done internally in large businesses has now been entirely subcontracted to other firms on a permanent basis, and not through a staffing agency, third-party management company, or platform. In these arrangements, the workers may be “traditional employees,” but they have been affected by an earlier outsourcing decision. It is important for policy makers to be able to track such trends in the use of subcontracting, which may involve W-2 employment but that “operate under very different economic constraints and incentives than had those jobs remained inside their original organizations” (Weil, 2019). Although such trends would not be captured within the question structure of the CWS, they certainly are important in measuring changes in the contracting relationships between companies and to the overall labor market landscape of compensation, benefits, and work conditions.
Business data also are needed to understand broader changes that have taken place in recent decades, whereby economic activity has been increasingly dominated by a select group of highly productive and profitable firms. Despite sluggish aggregate productivity growth, these leading firms have continued to experience steady growth in their productivity and financial returns (Furman and Orszag, 2018). As a result, industry sales are increasingly concentrated in firms with higher productivity. Correspondingly, the share of sectoral income arising from wages (the labor share) has been decreasing (Autor et al., 2017). Thus, as leading firms pull away from the pack of other companies in terms of revenue, jobs are not following them at the same pace, leading to diverging fortunes for workers who work in those firms as compared to those who do not.
20 See discussion of this problem in Weil (2019).
Meanwhile, other analyses have found that growth in earnings inequality can be primarily explained by growing differences between firms rather than within firms (e.g., Barth et al., 2016; Song et al., 2019). In particular, the firms that pay the most are becoming less likely to hire low-wage workers (Song et al., 2019). At the same time, these studies find that very large firms are less likely to be the ones that pay the most. One common view is that these trends primarily reflect the growing economic dominance of “superstar” firms, which increasingly differ from other firms in their sectors in their efficiency, technological sophistication, and dynamism (Autor et al., 2017). In this view, recent technological innovations, particularly in the digital economy, have enabled top firms to reap the benefits of scale to outstrip the efficiency of their competitors. As these firms learn to do more with less, they increase their productivity and capital returns but reduce their reliance on workers.
A critical question is whether the various AWAs and related changes in business structure may contribute to explaining the same set of facts. In this alternative view, these changes are not simply the result of changes in the technology used in production tasks; rather, they largely reflect “fissuring,” that is, the shifting of tasks performed within the firm to other parties through outsourcing or related forms of contracting (Weil, 2014). The rise in measured profitability and productivity of lead firms reflects not only that these firms are getting better at what they do, but also that they are making changes in choosing which functions to carry out in-house. As firms focus on core activities while shifting other activities to outside businesses within their industries or in other sectors, lead firms may improve their productivity and profitability on paper without fundamentally changing the work being done.
Additional data are needed to understand the extent to which the observed divergences between firms directly reflect the standard “superstar” view, in which technology contributes to observed changes (such as by providing lower-cost means of monitoring the performance of other entities) or the fissuring view. In the fissuring view, rising profits and productivity at lead firms may be fully decoupled from trends in aggregate productivity growth if firm boundaries change without major changes in overall activity. Growth policies for the economy geared to capital investment would play out differently depending on which of the two views, and therefore scenarios, is more accurate. Whereas the superstar story would imply a tradeoff of greater productivity enhancement but increased inequality, the fissuring story would imply that such policies can be expected simply to result in greater inequality with little benefit to overall growth.
This page intentionally left blank.